New AMD AM2 socket makes 939 heatsinks incompatible :(

All about them.

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

Mats
Posts: 3044
Joined: Fri Dec 26, 2003 6:54 am
Location: Sweden

Post by Mats » Wed Feb 08, 2006 2:00 pm

Maybe keeping it on topic, guys and girls? I didn't expect this kind of Spanish inquisition just because I told you what the Cell was...
Image

qviri
Posts: 2465
Joined: Tue May 24, 2005 8:22 pm
Location: Berlin
Contact:

Post by qviri » Wed Feb 08, 2006 2:10 pm

Mats wrote:I didn't expect this kind of Spanish inquisition just because I told you what the Cell was...
Nobody expects the Spanish Inquisition ;)

mellon
Posts: 105
Joined: Thu Apr 21, 2005 12:17 am
Location: Helsinki, Finland

Post by mellon » Wed Feb 08, 2006 2:27 pm

jaganath wrote: Can you give me a remotely plausible reason why a PhD would lie in a presentation intended solely for University of Texas electronics students? :roll:
Whee, that's a really difficult one! I mean seriously, it's obvious that the guy who has been working on the Cell and who is working for IBM would surely tell the absolute objective truth to such a selected and tight-lipped audience! And to make it better that information will never be disclosed publicly so it has to be the truth! </sarcasm>

Of course, here we are debating over a piece of publicly available presentation material. There is no such thing as a "presentation intended solely for <insert group here>" unless it is held behind closed doors with NDAs etc.

This honorable PhD would most likely not focus on telling us how it is difficult to develop parallel code that is reliable and still uses the Cell efficiently. It is even more likely that he would over-emphasize the possibilities of a grand future for Cell as his own work is somewhat dependant on the Cell development budget.

Overall, remain critical of the sources.

jaganath
Posts: 5085
Joined: Tue Sep 20, 2005 6:55 am
Location: UK

Post by jaganath » Wed Feb 08, 2006 2:48 pm

OK, so he's got a vested interest; anyone at this stage likely to be in possession of remotely relevant technical specs of the Cell core will almost certainly fall into this category. However, all this "oh he could be lying about it to save his own ass" is too "gunman on the grassy knoll" for me. Anyone clever enough to work on such a project should also realise that the product will stand or fall on its own merits, regardless of whatever futile efforts he may or may not make to hype it to a bunch of college grads.

So my problem with the conspiracy theory is:

A ) it insults his intelligence

B ) it insults our intelligence for thinking that we will fall for it and/or that we would not be appropriately sceptical of new information.

Put it another way: can you disprove that the Cell will have the characteristics that he claims it will have?

Elixer
Posts: 520
Joined: Wed Jun 02, 2004 10:31 am
Location: Las Cruces, NM
Contact:

Post by Elixer » Wed Feb 08, 2006 2:53 pm

jaganath wrote: Can you give me a remotely plausible reason why a PhD would lie in a presentation intended solely for University of Texas electronics students?
Not lie per say, just be full of shit. (couph couph couph ahem) Having grown up in Los Alamos (one of the highest PhDs per capita in the world) I can tell you that PhDs tend to rely on these things called theories. Unfortunately they tend to be blindsighted to the fact that things don't always behave as they do on paper.

Looking however at what is given, it looks good for the most part. 4Ghz might be a few years away, and that to me is very believable.

Tibors
Patron of SPCR
Posts: 2674
Joined: Sun Jul 04, 2004 6:07 am
Location: Houten, The Netherlands, Europe

Post by Tibors » Wed Feb 08, 2006 3:04 pm

There is a simple reason we won't see the Cell on the desktop. It is called legacy software. MS (and all other companies) won't go and rewrite their software for this non-x86 processor. So it would be limited to Linux and therefore (alas) not commercially viable.

On the other hand, the basic idea of the Cell is already seen on the desktop. The Cell is a fast bus to which are connected: a memory controller, a general purpose core and some specialised cores. Now look at the internal organisation of the A64. It is a fast bus to which are connected: a memory controller, a general purpose core, another general purpose core (later this year) and another two general purpose cores (sometime next year). There are also rumours they are going to connect a PCI-E controller directly on this bus.

Live
Posts: 101
Joined: Wed Nov 17, 2004 4:09 pm
Location: Uppsala, Sweden

Post by Live » Wed Feb 08, 2006 3:16 pm

jmke wrote: check out the mounting; all coolers which use the CLIPS system like the stock cooling (Freezer / HR-01 / etc..) will work fine with the new socket.

also I won't be too surprised to see a slightly changed mounting bracket which allows all others coolers to be installed on the new AMD CPU too.
AnandTech wrote:While the spacing width of the center mounting lug is the same on both the current cage and the AM2 version, that's about all that is the same. Most current basic designs connect with the center lug and have push down cam locks that lock onto the extra lug near the cage corners. There are no extra lugs on the new AM2 cage, so designs that require cam locking won't work. Neither will most of the "step-up" 3 lug mounting clips. AM2 only has one lug per side, and the 3-lug clip is normally intruding on the new screw locations (just a little too wide) to mount on the new AM2 cage.
Maybe I'm reading it wrong but it seems to me that AT disagrees with you.

Mats
Posts: 3044
Joined: Fri Dec 26, 2003 6:54 am
Location: Sweden

Post by Mats » Wed Feb 08, 2006 3:25 pm

Nevermind, a new bracket that fits old HS's and new mobos wouldn't cost much.



"...our primary weapons are fear and surprise! No, wait..."

Tibors
Patron of SPCR
Posts: 2674
Joined: Sun Jul 04, 2004 6:07 am
Location: Houten, The Netherlands, Europe

Post by Tibors » Wed Feb 08, 2006 5:16 pm

AnandTech wrote:Neither will most of the "step-up" 3 lug mounting clips. AM2 only has one lug per side, and the 3-lug clip is normally intruding on the new screw locations (just a little too wide) to mount on the new AM2 cage.
Why do you think there is a sticky thread about tin snips? They can fix (almost) everything :twisted:

jaganath
Posts: 5085
Joined: Tue Sep 20, 2005 6:55 am
Location: UK

Post by jaganath » Wed Feb 08, 2006 6:45 pm

"...our primary weapons are fear and surprise! No, wait..."
:lol:

Don't forget a fanatical devotion to the Pope and fluffy cushions.

BlueCan
*Lifetime Patron*
Posts: 74
Joined: Mon Oct 25, 2004 12:53 pm
Location: Newcastle, CA

Post by BlueCan » Wed Feb 08, 2006 6:52 pm

Tibors wrote:There is a simple reason we won't see the Cell on the desktop. It is called legacy software. MS (and all other companies) won't go and rewrite their software for this non-x86 processor. So it would be limited to Linux and therefore (alas) not commercially viable.
Good point! ...and that's the reason we'll never see MacOS on an Intel platform.

Oh. Wait.

-Patrick

Tibors
Patron of SPCR
Posts: 2674
Joined: Sun Jul 04, 2004 6:07 am
Location: Houten, The Netherlands, Europe

Post by Tibors » Wed Feb 08, 2006 8:17 pm

Apple earns their money by selling hardware, software is just a by product. There were reasons in the hardware division to switch. The relationship between Apple and IBM has been strained for some years. Now they just force the others to follow on their platform.

The only important hardware sales for MS is the Xbox, the competitor of..... On some other fronts (selling services and sofware) MS and IBM are competitors. MS and Intel have been been best buddies for years. Can you see any reason they want to pump huge amounts of money in Windows for Cell?

OS X is based on BSD. BSD for x86 has been availlable for years. The Cell is a variation on the PPC. Linux for PPC has been availlable for years too. So in both cases it was relative easy to create an OS. Windows for PPC? Don't know about you, but I have never seen it.

mellon
Posts: 105
Joined: Thu Apr 21, 2005 12:17 am
Location: Helsinki, Finland

Post by mellon » Thu Feb 09, 2006 12:39 am

jaganath wrote: So my problem with the conspiracy theory is:

A ) it insults his intelligence

B ) it insults our intelligence for thinking that we will fall for it and/or that we would not be appropriately sceptical of new information.

Put it another way: can you disprove that the Cell will have the characteristics that he claims it will have?
As this is a bit OT in this thread I'll keep it short. You can choose believe whatever he says but that is somewhat naïve. IBM has been promising fast speeds for its PPC line before and failed to keep those promises. Sony has been overhyping the performace of both previous Playstations.

There is absolutely no way for me to disprove anything as the Cell is physically not available to consumers in any form but that really doesn't mean that I should believe anything that is said about Cell automatically. You don't believe everything said in TV as-is even if you don't have the resources to prove them wrong do you?

I hope this didn't insult your intelligence too much :?

Erssa
Posts: 1421
Joined: Sat Mar 12, 2005 9:26 pm
Location: Finland

Post by Erssa » Thu Feb 09, 2006 3:14 am

mellon wrote:
jaganath wrote: So my problem with the conspiracy theory is:

A ) it insults his intelligence

B ) it insults our intelligence for thinking that we will fall for it and/or that we would not be appropriately sceptical of new information.

Put it another way: can you disprove that the Cell will have the characteristics that he claims it will have?
As this is a bit OT in this thread I'll keep it short. You can choose believe whatever he says but that is somewhat naïve. IBM has been promising fast speeds for its PPC line before and failed to keep those promises. Sony has been overhyping the performace of both previous Playstations.

There is absolutely no way for me to disprove anything as the Cell is physically not available to consumers in any form but that really doesn't mean that I should believe anything that is said about Cell automatically. You don't believe everything said in TV as-is even if you don't have the resources to prove them wrong do you?

I hope this didn't insult your intelligence too much :?
Hear hear!
Agreed 100%

jaganath
Posts: 5085
Joined: Tue Sep 20, 2005 6:55 am
Location: UK

Post by jaganath » Thu Feb 09, 2006 5:31 am

I guess the difference is I want to believe it's true and you guys don't. Why that is I couldn't possibly say....

Mr Evil
Posts: 566
Joined: Fri Jan 20, 2006 10:12 am
Location: UK
Contact:

Post by Mr Evil » Thu Feb 09, 2006 7:13 am

jaganath wrote:I guess the difference is I want to believe it's true and you guys don't. Why that is I couldn't possibly say....
Years of hearing about marvelous technologies that end up not being very impressive at all. It's extremely rare that new hardware lives up to the hype, if it even materializes at all. I'm sure we all want it to be true, but some are more jaded than others. Now where did I put my VR goggles..?

Mats
Posts: 3044
Joined: Fri Dec 26, 2003 6:54 am
Location: Sweden

Post by Mats » Thu Feb 09, 2006 7:56 am

Mr Evil wrote:Now where did I put my VR goggles..?
:lol: :lol: :roll: :lol:
I'd almost forgotten them.

ronrem
Posts: 1066
Joined: Sun Jan 16, 2005 2:59 am
Location: Santa Cruz

Post by ronrem » Thu Feb 09, 2006 8:15 am

McBanjo wrote:
anaqer wrote:
McBanjo wrote:Why buy a new one if you have a good old one?
Because the 'good old one' may not be good enough anymore...? Keeping in mind the (supposedly quite a bit) higher TDPs for the M2 processors, I don't think you'll really want to hold on to whatever you have at the moment.
Then it isn't good old one, just old one ;-)
Assuming it still works and still is good. Doubt Ninja will have trouble handling that extra heat for example.
Not sure TDP will go up since they'll be using a smaller manufactoring technic (45nm?). I'm guessing tho. It'll probibly go up some but that might cancel out some of the raise of TDP
jaganath wrote:Why would you get an M2 processor if the TDP is worse and the performance not much better than current S939 procs?
Performance in the CPU will rise and you'll be using DDR2 so thats some performance as well. Don't know if you gain anything else in the socket itself. Probibly tho
This is the SILENT forum...not te OVERCLOCK TILL IT GLOWS forum. Unless I'm gonna play Doom for money,I'm not sure I'm doing anything that justifies the power difference between a Venice 3200 and an X2 4600. I don't type fast on the Keyboard,and my brain can't keep up with an old Celeron. I guess it would be cool to convert an 800 mb audio file 2 minutes faster but that's when I go to the kitchen for a snack,or a coffee.

If M2 does not use the 65nm process for LESS heat,but only for even more speed overkill...but heat/noise worse than a Prescott, I'd be doing my next rig with a plain Sempron.

I added simple brackets so my socket A sin could use the mobo's bolt holes instead of a funky clip I didn't trust. I doubt there are many heatsinks for 754/939 that can't be modde to fit a 4 bolt mount. If there was no bolt hole option,just clips,then modding would be tough.

ronrem
Posts: 1066
Joined: Sun Jan 16, 2005 2:59 am
Location: Santa Cruz

Post by ronrem » Thu Feb 09, 2006 8:29 am

I'd heard there is a future Sempron on the 939. The current semprons replaced the mainstream Newcastle on 754 when Winchester became the new 90 nm chip on 939. Semprons soon got 90nm (Palermo) the cool runners of the AMD line. When the mainstream moves on to M2,Sempron will graduate to 939 and you can bet there will be a process shrink to 65nm before long. That is probably where I'll find silent happiness. I'd guess there will be an under cached Sempy 3500 or such that is a 65 nm item,with a price similar to todays 3000 Palermo and even a bit less heat output. Such a chip won't need a fan on a good heatsink unless overclocked. Getting Quiet will get even easier.

Mats
Posts: 3044
Joined: Fri Dec 26, 2003 6:54 am
Location: Sweden

Post by Mats » Thu Feb 09, 2006 8:58 am

ronrem wrote:If M2 does not use the 65nm process for LESS heat,but only for even more speed overkill...but heat/noise worse than a Prescott, I'd be doing my next rig with a plain Sempron.
AM2 is not equal to 65 nm process, it will take a year before we'll see 65 nm CPU's showing up from AMD. Until then AMD need headroom for upcoming 2.67 and 3.00 GHz dual core FX CPU's, simple as that. It doesn't mean that TDP must go up for every single model compared to the S939 counterparts, even though a possible faster HTT might need a little extra juice (still don't know if it will be 200, 333, or 400 MHz).

rpsgc
Friend of SPCR
Posts: 1630
Joined: Tue Oct 05, 2004 1:59 am
Location: Portugal

Post by rpsgc » Thu Feb 09, 2006 9:34 am

ronrem wrote:I'd heard there is a future Sempron on the 939.
Welcome to last year(?). OEM only though.
ronrem wrote:When the mainstream moves on to M2,Sempron will graduate to 939 and you can bet there will be a process shrink to 65nm before long.
Or, shall I say, no. When M2 launches there'll be M2 Semprons too. 65nm shrink only in late 2006 or 2007. Way after M2 launches.

anaqer
Posts: 44
Joined: Sun Jun 26, 2005 11:31 pm

Post by anaqer » Thu Feb 09, 2006 10:50 am

jaganath wrote:OK, I know what you're really trying to say is "It's cr@p and I already know why", so why don't you go ahead and explain why the Cell will be a disappointment rather than this kind of cryptic comment.
Not outright 'crap', just woefully inadequate for the desktop - if you want the details, I suggest you hit Ars or Anand for last year's write-ups on Cell. The gist of it is that as good as it may be for certain tasks, you just don't want to run Windows & Co. on a heavily FPU-biased inorder processor. (A revised Cell 'Lite' may be ok for media centers though.)

TooNice
Posts: 98
Joined: Mon Feb 06, 2006 12:57 pm

Post by TooNice » Mon Feb 27, 2006 8:22 pm

TooNice wrote:Umm, I am building a new system. As much as I am tempted by the Scythe Ninja, would it be better if I stick with the stock cooler (since I am building a retail chip anyway), until newer coolers/AM2 compatible revisions of the coolers are released?).
Bump and going back to this... I am still not decided. Unlike the US, the Ninja is not inexpensive over here (i.e. it cost almost as much as the Zalman 9500 before you add a Yate Loon/Nexus to it).

I consider myself to have zero modding ability and the only tool I have is a set of screwdrivers. If I am not mistaken, a new Scythe product is on the way. So would it be better to just live with a stock cooler until then? I assume that the new Scythe will be designed to be AM2 ready, AND backward compatible with socket 939. I am also kind of assuming that the AM2 HS mounting system will last for a while, perhaps beyond AM2 (in other word, I expect a good AM2 HS to be re-usable for whatever comes after that).

Post Reply